Diagnosis of Alcoholism in a Simulated Patient Encounter by Primary Care Physicians

Richard L. Brown, MD, MPH, William B. Carter, PhD, and Michael J. Gordon, PhD Philadelphia, Pennsylvania, and Seattle, Washington

Although early detection and treatment of alcoholism have been shown to be effi cacious, it is widely accepted that primary care physicians often fail to diagnose alcoholism. In this study, a computerized, simulated encounter with an alcoholic patient was used to assess the performance of a randomly selected sample of pri mary care physicians in diagnosing alcoholism. Of 95 physicians in this study, only 32 percent diagnosed alcoholism with maxi mal certainty. There was great variability among physicians in the threshold of in formation needed to diagnose alcoholism. One third of subjects misinterpreted symptoms of alcoholism and erroneously made other psychiatric diagnoses, chiefly anxiety or depression. Results of this pilot study were not associated with the physicians’ age, sex, specialty, duration of training, or reported personal impact of alcoholism. This study provides further evidence of the need for additional education of primary care physicians if such physicians are to succeed in reducing the dramatic im pact of alcoholism and alcohol abuse on public health.

lcoholism and alcohol abuse are among the most have a sizable opportunity to improve the health outcomes A prevalent and serious public health problems in the of a large segment of their patient populations. nation, with 14 percent of the adult population estimated This study was an initial attempt to assess the perfor to be “symptomatic drinkers.”1 In 1980 the total mortality mance of practicing primary care physicians in diagnosing attributed to alcoholism in the United States was over early alcoholism by using a computerized, simulated pa 69,000.1 Additionally, alcohol abuse contributes substan tient encounter. It was intended that the results would tially to many morbid conditions,2 is linked to family dis provide an estimate of the frequency with which primary ruption, child abuse, and violence,2 3 and is estimated to care physicians might diagnose alcoholism in patients cost the nation as much as $116.7 billion yearly.1 similar to the simulated patient and, more important, that Alcoholism can be diagnosed before irreversible they would provide insight into aspects of physicians’ biomedical consequences have occurred,4-6 and, once di thought processes, which, if altered, might facilitate phy agnosed, it can be successfully treated.7-10 Nevertheless, sicians’ diagnosing of alcoholism. physicians often fail to diagnose alcohol problems,11-14 and when such problems are diagnosed, physicians often fail to attempt to treat them.14-18 Although from 10 to 40 METHODS percent of patients seen by physicians have alcohol prob lems,19 only 15 percent of all alcoholics are believed ever Subjects and Recruitment to receive treatment.20 Primary care physicians therefore Recruiting letters were mailed to a random sample of the approximately 400 board-certified family physicians and

Submitted, revised, June 23, 1987. general internists (without subspecialty board certification) who were included on the mailing lists of the King County From the Department of Family Medicine, University of Washington, Seattle, (Washington) Medical Society. The letters offered two Washington, and the Department of Family Medicine, Thomas Jefferson University, Philadelphia, Pennsylvania. This study was performed while Dr. Richard Brown category I continuing medical education (CME) credits, was a Robert Wood Johnson Fellow in Family Medicine, University of Washington, $25, and educational feedback in return for completing Seattle, Washington. Requests for reprints should be addressed to Dr. Richard i Brown, Department of Family Medicine, Thomas Jefferson University, 1015 a computerized, simulated patient encounter and a follow Walnut Street, Suite 401, Philadelphia, PA 19107. up questionnaire. Physicians were informed that the

TABLE 1. CLUES TO THE DIAGNOSIS OF ALCOHOLISM IN TABLE 2. CLUES TO THE DIAGNOSIS OF ALCOHOLISM ON THE SIMULATED PATIENT THE ALCOHOL USE MENU

1. Drinking in response to stress 1. Admits greater alcohol consumption than previously 2. Marital discord 2. History of alcoholic blackouts 3. Difficulties at work 3. Previous attempts to decrease drinking 4. Recent change of job 4. Wife’s concern about patient’s drinking 5. History of a one-car motor vehicle accident 5. Patient’s guilt over his drinking 6. Paternal history of alcoholism 6. Morning consumption of alcohol 7. Failure to reduce coffee intake 8. Failure to reduce cigarette use 9. Feelings of discouragement about his situation 10. Sleep disturbance 11. No appetite disturbance appropriate certainty and to formulate a complete man 12. Slight weight gain agement plan. Each physician then performed a diagnostic 13. Slightly disturbed attention and concentration evaluation of the same simulated patient by means of a 14. Normal libido menu-driven computer program that was created for this 15. Slightly fatigued study. The menus presented physicians with over 600 de 16. No suicidal thoughts 17. No previous psychiatric history scriptions of pieces of clinical data that physicians could 18. Anxious feelings about work request on the patient. In so doing, physicians performed 19. Intermittently elevated blood pressure a history and physical examination, ordered and received 20. Blood alcohol level of 9 mmol/L (40 mg/dL) laboratory tests, and reviewed past medical records. The 21. Serum glutamic-oxaloacetic transaminase (SGOT) of 0.09 Mkat/L (54 U/L) computer program recorded the order in which each sub 22. Gamma glutamyltransferase of 5.07 /*kat/L (304 U/L) ject obtained various pieces of clinical data. Almost all 23. High-normal mean corpuscular volume physicians completed the exercise within 45 minutes. 24. Gastritis visible on gastroscopy The case was one of a 38-year-old male, married, in ventory control manager with recurrent abdominal pain. If appropriate items were selected, the history revealed symptoms of active peptic ulcer and gastritis and many study’s purpose would be to investigate how physicians clues to alcoholism,22,23 as shown in Table 1. Eighteen of would manage a particular patient in their practices. Full these clues were strictly historical, including the patient’s time physicians at the authors’ institution and close ac admission of drinking in response to stress. Several of quaintances of the authors were excluded. Ninety-five these historical clues, such as anxiety related specifically physicians (39 percent of those contacted) were enrolled to the patient’s job and a normal libido, were designed to in the study. More family physicians were recruited and help rule out a diagnosis of anxiety disorder or depression. enrolled because the supply of general internists was ex Five clues were laboratory results, and one, high blood hausted. Differences between the participation rates of pressure, was obtainable by history, physical examination, family physicians and general internists were not statis and past medical records. tically significant. Throughout the exercise, the assistant encouraged the physicians to request additional information not found on the computer menus. Three additional menus dealing The Computerized, Simulated with marital and family problems, stress, and alcohol use Patient Encounter were made available by the assistant if physicians ex Initially, the computerized, simulated case was pretested pressed an interest in these specific areas. Responses to on six faculty members, including two with special ex items on the alcohol-use menu provided further infor pertise in alcoholism. The case fulfilled the Diagnostic mation on the extent of the patient’s drinking problem and Statistical Manual of Mental Disorders, Third Edition (Table 2). (DSM III) criteria for a diagnosis of alcohol abuse but not Immediately after completing the exercise, subjects for any other psychiatric diagnosis based on DSM III cri were given a questionnaire on which they indicated the teria.21 Other correct diagnoses were gastritis and peptic likelihood that the patient had any of 11 diagnoses—gall ulcer. All six faculty members found the case to be realistic bladder disease, gastritis, gastrointestinal tumor, pancre and the correct diagnoses to be accurate. atitis, peptic ulcer, alcohol problem, anxiety disorder, A trained research assistant administered the simulation depression, malingering, personality disorder, and so to each physician on a portable microcomputer, usually matization disorder— on a five-point Likert-type scale: in the physician’s office. Each physician was instructed to very likely, somewhat likely, intermediate, somewhat un obtain sufficient information to make a diagnosis with likely, and very unlikely.

260 THE JOURNAL OF FAMILY PRACTICE, VOL. 25, NO. 3, 1987 DIAGNOSIS o f a lc o h o lis m

TABLE 3. PERCENTAGE OF SUBJECTS WHO SELECTED PSYCHOSOCIAL DIAGNOSES ON THE SIMULATED PATIENT

Very Somewhat Somewhat Very Diagnosis Likely Likely Intermediate Unlikely Unlikely Missing

Alcohol problem 32 33 17 15 4 0 Anxiety d iso rd e r 28 35 24 11 0 2 Depression 23 44 19 11 3 0 Malingering 0 0 1 17 79 3 Personality disorder 3 7 16 24 46 3 Somatization disorder 4 8 5 26 53 2

Follow-up Questionnaire TABLE 4. DIAGNOSES OF ALCOHOL PROBLEM, ANXIETY Several weeks after completing the computerized case, 93 DISORDER, AND DEPRESSION MADE ON THE percent of the subjects completed a questionnaire de SIMULATED PATIENT scribing their medical training, their office practices, and Anxiety or the impact of alcoholism on their personal lives. Anxiety or Depression* Depression* Not Diagnosed Diagnosed Totals

RESULTS Alcohol problem** diagnosed 11 19 30

Subjects Alcohol problem** not diagnosed 25 40 65 Of all 95 subjects, 54 percent were family physicians, 44 percent were general internists, and 2 percent were board Totals 36 59 95 certified in both specialties. Their mean age was 42.3 years * Incorrect diagnosis (standard deviation 10.3 years). Ten (11 percent) were * * Correct diagnosis female. Of the 88 who returned the follow-up question naire, 56 percent had completed exactly three years of postgraduate training, 30 percent four years or more, and likely. Subjects’ performance in assessing the biomedical 15 percent less than three years. One half of these subjects diagnoses, plus the observation that all subjects made a indicated that drinking problems have had at least “some “very likely” response for at least one diagnosis, indicated impact on their personal lives, either through family, that each subject used a “very likely” response to indicate friends, or themselves.” Subjects reported seeing from 12 a certain, principal diagnosis. to 200 (mean 83.6) ambulatory patients in 6 to 80 (mean 40.7) hours each week. Clues Subjects uncovered an average of 11.2 (standard deviation Diagnoses 4.7) of the 19 nonlaboratory clues to alcoholism. Although A total of 30 subjects (32 percent) diagnosed alcoholism the total number of nonlaboratory clues uncovered was as very likely. Of these, 11 also diagnosed at least one not associated with the certainty of diagnosis of alcohoL other psychosocial condition, chiefly anxiety or depres ism, the uncovering of two of these clues was associated sion, which by DSM III criteria would be inappropriate with rating an alcohol problem as very likely according diagnoses (Table 3).21 Twenty-six (27 percent) of the sub to chi-square tests. These clues were (1) a previous history jects erroneously diagnosed one or more psychiatric of a one-car motor vehicle accident (P < .01) and (2) a problems (usually anxiety or depression [Table 4], but paternal history of alcoholism (P = .05). Discoveries of a also somatization and personality disorders) and did not mean corpuscular volume (MCV) at the upper end of the diagnose an alcohol problem. Thirty-nine (41 percent) normal range and of gastritis on gastroscopy were not identified no psychosocial diagnosis to be very likely. In related to diagnosing the alcohol problem. Laboratory contrast, at least one of the two correct biomedical di findings of three tests were clues associated with the di agnoses, peptic ulcer and gastritis, was rated as very likely agnosis of an alcohol problem: a blood alcohol level of 9 by 91 (96 percent) subjects, and only one rated another mmol/L (40 mg/dL), elevated serum glutamic-oxaloacetic biomedical diagnosis (gastrointestinal tumor) as very transaminase, and elevated gamma-glutamyltransferase.

the JOURNAL OF FAMILY PRACTICE, VOL. 25, NO. 3, 1987 261 DIAGNOSIS OF ALCOHOLISM

specialty, duration of training, or reported personal impact of alcoholism.

DISCUSSION

It seems clear that subjects diagnosed the psychosocial problem, namely alcoholism, with much less accuracy and certainty than the biomedical problems, peptic ulcer and gastritis. More than 40 percent did not make any psychosocial diagnosis at all. Most interesting, however, were the vastly different informational thresholds among physicians in diagnosing alcoholism. Some physicians di agnosed alcoholism on the basis of only general historical clues, while others who possessed information on elevated liver function tests or the results of a detailed alcohol his Figure 1. Paths taken by physicians in the simulation ex tory did not make the diagnosis. ercise leading to their assessments of the likelihood of an Several methodological issues deserve comment. It is alcohol problem. Numbers show how many subjects, of a questionable how the diagnoses that physicians indicated total of 95, took the respective paths on the likelihood scale would relate to those that the phy sicians would make in a more open-ended fashion. For example, does a “very likely” response truly indicate a certain diagnosis, and does a “somewhat likely” response Paths truly indicate a less than certain diagnosis? This study used the responses to questions on biomedical diagnoses Paths that subjects took relative to their assessments of as a guide to interpreting physicians’ responses to similar alcoholism are diagrammed in Figure 1. Based on infor questions on psychosocial diagnoses. Unfortunately, it is mation gathered by general history and physical exami not known whether physicians judge the certainty of psy nation, four subjects requested alcohol history beyond that chosocial diagnoses in the same way they judge that of which was immediately available on the menus. Two of biomedical diagnoses. It seems reasonable, however, that these then uncovered the elevated liver function tests and a physician’s failure to indicate a maximal degree of cer diagnosed alcoholism. The other two did not obtain these tainty about a diagnosis of alcoholism would predict a laboratory tests and rated alcoholism as only somewhat failure to provide effective treatment for it, but there are likely. Fifty-eight subjects found elevated liver function no data to support this. tests without previously having requested additional al The use of a computerized, simulated patient encounter cohol-related history. Only nine of these then went on to in the study raises questions of validity. To what degree request additional alcohol-related history, and eight of do the subject’s performances on this computerized, sim these nine diagnosed alcoholism. Forty-nine of the 58 ulated patient encounter correspond to the frequency with subjects (84 percent) who obtained liver function tests which and the manner in which the subjects would di never sought additional alcohol-related history. Of these agnose early alcoholism in clinical practice? The lack of 49 subjects, 15 subsequently diagnosed alcoholism, and nonverbal cues from the simulated patient may have hin 16 rated it less than somewhat likely. The remaining 33 dered physicians in performing psychosocial evaluations. subjects neither found abnormal liver function tests nor On the other hand, physicians may have been less inhib requested additional alcohol-related history. Five of these ited about asking personal questions of the simulated pa nevertheless diagnosed alcoholism as very likely. tient than of humans, who can exhibit or elicit even greater In summary, 15 percent of those who obtained neither anger and discomfort than computers. Cueing effects, liver function tests nor additional alcohol history had suf produced by listing of possible pieces of clinical data that ficient information to diagnose alcoholism with maximal could be obtained on the simulated patient, may have certainty. However, 62 percent of the 60 who had uncov provoked physicians to choose options that they would ered elevated liver function tests and 23 percent of the 13 not actually perform in clinical practice.24'27 Relevant who had obtained a detailed positive alcohol history did cueing was minimized in this study by including over 600 not rate alcoholism as a very likely diagnosis. items and by revealing more detailed menus on alcohol None of the results, whether relating to diagnoses, clues, use, stress, and family or marital problems only in re or paths to diagnosis, were associated with age, gender, sponse to specific requests. In any case, cueing problems

262 THE JOURNAL OF FAMILY PRACTICE, VOL. 25, NO. 3,1987 diagnosis o f a lc o h o lis m

are thought to bias results in such a manner that clinicians Unfortunately, even optimization of primary care phy perform better on patient management problems than sicians’ diagnosis of alcoholism would be insufficient, as they actually do in clinical practice.27 one recent study showed that for many patients who were Other studies have explored aspects of the validity of diagnosed as alcoholic, no attempts at treatment were other simulated patient encounters. Investigators have made.14 Clearly, further education of primary care phy found significant relationships between performance on sicians is necessary if such physicians are to have more patient management problems and scores on more tra of an impact on reducing the public health impact of al ditional multiple-choice examinations.28"31 In addition, coholism and alcohol abuse. experienced clinicians outperform residents, who, in turn, surpass medical students in performance on patient man Acknowledgment agement problems.24,30,31 Nevertheless, minor variations in the content of the case might have produced different Financial support for this study was provided by the Family Health Foundation of America, and the J.M. Foundation. results.32 Further assessment of the validity of this study would require additional investigation. Generalizability of the study’s results to all primary References care physicians of King County would require further 1. Harwood HJ, Napolitano DM, Kristiansen PL, et al: Economic characterization of nonparticipants. Some comfort with Costs to Society of Alcohol and Drug Abuse and Mental Illness: generalizability may be derived from the fact that physi 1980. Publication RTI/2734/00-01FR. Research Triangle Park, cians were not aware of the focus of the study when they NC, Research Triangle Institute, 1984 2. West LJ, Maxwell DS, Noble EP, et al: Alcoholism. Ann Intern decided whether to participate. Also, data on the subjects’ Med 1984; 100:405-416 practices demonstrate that many busy physicians partic 3. Institute of Medicine, Division of Health Promotion and Disease ipated in this study. Prevention: Alcoholism, Alcohol Abuse, and Related Problems: This study suggests that many primary care physicians Opportunities for Research. Washington, DC, National Academy of Sciences, 1980 may fail to diagnose early alcoholism, despite the avail 4. Bernadt MW, Mumford J, Taylor C, et al: Comparison of ques ability of many symptoms and supporting laboratory ev tionnaire and laboratory tests in the detection of excessive drinking idence, and that there may be marked differences among and alcoholism. Lancet 1982; 1:325-328 physicians in informational thresholds for diagnosing al 5. Peterson B, Treel E, Kristianson H: Comparison of gamma-glu coholism. Some physicians who miss the diagnosis, such tamyl transferase and questionnaire tests as alcohol abuse indi cators in different risk groups. Drug Alcohol Depend 1983; 11: as those who uncovered many historical clues and ab 279-286 normal laboratory tests, may fail to generate a hypothesis 6. Bernadt MW, Mumford J, Murray RM: A discriminant function of alcoholism. Others who miss the diagnosis, such as analysis of screening tests for excessive drinking and alcoholism. those who had received additional, pathognomonic, his J Stud Alcohol 1984; 45:81-86 7. Emrick CD: A review of psychologically oriented treatment of al torical information on alcohol use, may be reluctant to coholism: I. The use of interrelationships of outcome criteria and make what they consider to be a judgmental diagnosis, drinking behavior following treatment. Q J Stud Alcohol 1974; the treatment of which may demand an unpleasant con 35:523-549 frontation with the patient. Some physicians may be un 8. Emrick CD: A review of psychologically oriented treatment of al aware of the extent to which alcoholism and its psycho coholism: II. The relative effectiveness of treatment versus no treatment. J Stud Alcohol 1975; 36:88-108 social consequences to its victims can mimic or elicit 9. Baekeland F, Lundwall L, Kissin B: Methods for treatment of symptoms of anxiety or depression. Others, such as some chronic alcoholism: A critical appraisal. In Gibbins R, Israel Y, of those who made no principal psychosocial diagnosis, Kalant H, et al (eds): Research Advances in Alcohol and Drug may feel that doing so would be beyond the scope of their Problems. New York, John Wiley & Sons, 1975, pp 247-327 10. Brandsma JM, Maultsby MC Jr, Welsh RJ: Outpatient Treatment duties in providing primary care. of Alcoholism: A Review and Comparative Study. Baltimore, Uni The results of this study are consistent with those of versity Park Press, 1980 many others that show that physicians often miss the di 11. Fisher JV, Fisher JC, Mason RL: Physicians and alcoholics: Mod agnosis of alcoholism.11-14 In addition, this study supports ifying behavior and attitudes of family practice residents. J Stud the results of one particular study that suggest that primary Alcohol 1976; 37:1686-1693 12. Moore RD, Malitz FE: Underdiagnosis of alcoholism by residents care physicians may have difficulty in diagnosing alco in ambulatory medical practice. J Med Educ 1986; 61:46-52 holism with coexistent symptoms of depression and with 13. Coulehan JL, Zettler-Segal M, Block M, et al: Recognition of al out a coexistent antisocial personality disorder.13 coholism and substance abuse in primary care patients. Arch Intern There is one report of improvement in diagnosis of Med 1987; 147:349-352 14. Bush B, Shaw S, Cleary P, et al: Screening for alcohol abuse alcoholism following an educational intervention.11 Such using the CAGE questionnaire. Am J Med 1987; 82:231-235 improvement, however, occurred to a point at which 15. Hayman N: Current attitudes to alcoholism of psychiatrists in physicians had identified 2.5 percent of their patients as Southern California. Am J Psychiatry 1956; 112:484-493 alcoholics, at least fourfold less than would be expected. 16. Jones RW, Helrich AR: Treatment of alcoholism by physicians in

the JOURNAL OF FAMILY PRACTICE, VOL. 25, NO. 3, 1987 263 DIAGNOSIS OF ALCOHOLISM

private practice: A national survey. J Stud Alcohol 1972; 33:117- 25. Newbie Dl, Hoare J, Baxter A: Patient management problems1 131 Issues of validity. Med Educ 1982; 16:137-142 17. Caswell S, MacPherson M: Doctors and alcohol: The responses 26. Norman GR, Feightner JW: A comparison of behavior on simulated of a self-selected group of general practitioners to patients with patients and patient management problems. Med Educ 1981 • 15- alcohol-related problems. NZ Med J 1982; 95:462-466. 26 18. Anderson P: Managing alcohol problems in general practice. Br 27. Goran MJ, Williamson JW, Gonella JS: The validity of PMPs J Med J 1985; 290:1873-1876 Med Educ 1973; 43:171-177 19. Lewis DJ, Gordon AJ: Alcoholism and the general hospital: The 28. Norcini J, Swanson D, Webster G: Reliability, validity, and efficiency Roger Williams Intervention Program. Bull NY Acad Med 1983; of various item formats in assessment of physician competence. 59:181-197 Conf Res Med Educ Proc 1983; 22:53-58 20. Saxe E, Dougherty D, Esty K: The effectiveness and cost of al coholism treatment: A public policy perspective. In Mendelson JA 29. Hubbard JP: Objective evaluation of clinical competence, In Hub and Mello NK (eds): The Diagnosis and Treatment of Alcoholism, bard JP (ed): Measuring Medical Education. Philadelphia, Lea & ed 2. New York, McGraw-Hill, 1985, pp 485-539 Febiger, 1971 21. Diagnostic and Statistical Manual of Mental Disorders, ed 3 (DSM 30. Newbie Dl, Hoare J, Elmslie RG: The validity and reliability of a III). Washington, DC, American Psychiatric Association, 1980 new examination of clinical competence of medical students Med 22. Skinner HA, Holt S: Early intervention for alcohol problems. J R Educ 1981, 15:46-52 Coll Gen Pract 1983; 33:787-791 31. Mazzuca SA, Cohen SJ, Clark CM: Evaluating clinical knowledge 23. Manual on Alcoholism, ed 3. Monroe, Wise, American Medical across years of training. J Med Educ 1981; 56:83-90 Association, 1977 32. Norman GR, Feightner JW, Tugwell P, et al: The generalizability 24. Page GC, Fielding DW: Performance on PMPs and performance of measures of clinical problem-solving. Conf Res Med Educ Proc in practice: Are they related? J Med Educ 1980; 55:529-537 1983; 22:110-114

264 THE JOURNAL OF FAMILY PRACTICE, VOL. 25, NO. 3,1987